A Multi-Site Study of Radio Environment for Cosmology Experiments

TL;DR

This work addresses the challenge of radio frequency interference (RFI) in low-frequency cosmology by performing a portable, multi-site RFI survey across 30–300 MHz at four diverse sites (GRO, Twin Lakes, Kalpong Dam, and Gruvebadet). It employs two complementary detection methods—Hampel filtering for outliers and singular value decomposition (SVD) for broad-band patterns—to quantify RFI occupancy, identify persistent versus transient interference, and characterize spectral morphology. The analysis yields site-specific occupancies and broad-band RFI clusters, revealing that GM and UHF bands are severely affected at GRO, while Twin Lakes at night and Kalpong Dam offer comparatively cleaner wideband conditions; Gruvebadet shows strong low-frequency RFI but cleaner bands above ~100 MHz. The study demonstrates a generalizable RFI characterization framework and informs site selection and mitigation strategies for SARAS-like global 21-cm cosmology experiments, highlighting practical deployment options (notably TLL and KDA) and the importance of combining Hampel and SVD analyses to capture both narrow- and broad-band interference facets.

Abstract

Radio Frequency Interference (RFI) presents a significant challenge for carrying out precision measurements in radio astronomy. In particular, RFI can be a showstopper when looking for faint cosmological signals such as the red-shifted 21-cm line from cosmic dawn (CD) and epoch of reionization (EoR). As wireless communications, satellite transmissions, and other RF technologies proliferate globally, understanding the RFI landscape has become essential for site selection and data integrity. We present findings from RFI surveys conducted at four distinct locations: three locations in India, the Gauribidanur Radio Observatory in Karnataka, Twin Lakes in Ladakh, Kalpong Dam in the Andaman Islands, and the Gruvebadet Atmosphere Laboratory in Ny-Ålesund, Svalbard, Norway. These sites, selected based on their geographical diversity and varying levels of human activity, were studied to assess RFI presence in 30-300 MHz bands, critical for low-frequency observations and experiments targeting the 21-cm CD/EoR signal. Using an automated RFI detection approach via the Hampel filter and singular value decomposition, the surveys identified both persistent and transient interference, which varies with location and time. The results provide a comprehensive view of the RFI environment at each site, informing the feasibility of long-term cosmological observations and aiding in the mitigation of RFI in radio astronomical data. The methods developed to characterize RFI can be easily generalized to any location and experiment.

Figures (19)

• Figure 1: The top panel shows a block diagram of the RFI radiometer setup. The pre-amplifier module amplifies the weak RF signal from the discone antenna. The post-amplifier unit, RF analyzer, and data acquisition laptop are housed inside an RF-shielded enclosure (shown within a dashed box). Apart from pictures of the discone antenna and the RF-shielded enclosure, the bottom panel shows an open view of the internal circuitry of the pre-amplifier and post-amplifier unit.
• Figure 2: Geometry of the discone antenna as simulated in FEKO. The beam pattern shown is at 150 MHz.
• Figure 3: Comparison of the reflection coefficient of the antenna with simulations in FEKO: blue dashed line shows simulation with realistic ground profile, orange shows the measurement at GRO, and green represents measurement at GLS. The GRO profile is less noisy than GLS because of the integration times of measurements.
• Figure 4: The top panel shows the gain variation of the RF signal chain as a function of input power at the pre-amplifier input port. The bottom panel depicts the linearity plot of the RF signal chain. As the input power to the pre-amplifier module increases, the receiver chain moves from a linear regime to the 1 dB compression point of the amplifier in the post-amplifier unit. Both plots were obtained by sweeping the power of a tone at 150 MHz.
• Figure 5: Satellite images of the surveyed locations along with the horizon profile around it obtained using the Python package shapes. (a) GRO (b) TTL (c) KDA (d) GLS.